Pascal Mettes
@pascalmettes.bsky.social
Assistant Professor, University of Amsterdam
Hyperbolic deep learning
Hyperbolic deep learning
Reposted by Pascal Mettes
🚨🚨📄 Check out our new preprint!
Our results reveal novel insights on how continuous visual input is integrated in the human brain💡, beyond the standard temporal processing hierarchy from low to high-level representations
Our results reveal novel insights on how continuous visual input is integrated in the human brain💡, beyond the standard temporal processing hierarchy from low to high-level representations
📢 New preprint, together with @sargechris.bsky.social!
Building on @sargechris.bsky.social's previous work, we benchmark 100+ image and video models 🤖 on brain representational alignment, this time to EEG data of humans 🧠 watching videos! 🧵⬇️
www.biorxiv.org/content/10.1...
Building on @sargechris.bsky.social's previous work, we benchmark 100+ image and video models 🤖 on brain representational alignment, this time to EEG data of humans 🧠 watching videos! 🧵⬇️
www.biorxiv.org/content/10.1...
The Human Brain as a Dynamic Mixture of Expert Models in Video Understanding
The human brain is the most efficient and versatile system for processing dynamic visual input. By comparing representations from deep video models to brain activity, we can gain insights into mechani...
www.biorxiv.org
October 31, 2025 at 1:32 PM
🚨🚨📄 Check out our new preprint!
Our results reveal novel insights on how continuous visual input is integrated in the human brain💡, beyond the standard temporal processing hierarchy from low to high-level representations
Our results reveal novel insights on how continuous visual input is integrated in the human brain💡, beyond the standard temporal processing hierarchy from low to high-level representations
Reposted by Pascal Mettes
📢 New preprint, together with @sargechris.bsky.social!
Building on @sargechris.bsky.social's previous work, we benchmark 100+ image and video models 🤖 on brain representational alignment, this time to EEG data of humans 🧠 watching videos! 🧵⬇️
www.biorxiv.org/content/10.1...
Building on @sargechris.bsky.social's previous work, we benchmark 100+ image and video models 🤖 on brain representational alignment, this time to EEG data of humans 🧠 watching videos! 🧵⬇️
www.biorxiv.org/content/10.1...
The Human Brain as a Dynamic Mixture of Expert Models in Video Understanding
The human brain is the most efficient and versatile system for processing dynamic visual input. By comparing representations from deep video models to brain activity, we can gain insights into mechani...
www.biorxiv.org
October 31, 2025 at 1:03 PM
📢 New preprint, together with @sargechris.bsky.social!
Building on @sargechris.bsky.social's previous work, we benchmark 100+ image and video models 🤖 on brain representational alignment, this time to EEG data of humans 🧠 watching videos! 🧵⬇️
www.biorxiv.org/content/10.1...
Building on @sargechris.bsky.social's previous work, we benchmark 100+ image and video models 🤖 on brain representational alignment, this time to EEG data of humans 🧠 watching videos! 🧵⬇️
www.biorxiv.org/content/10.1...
Reposted by Pascal Mettes
We are thrilled to welcome Aiden Durrant as a Winter School speaker at NLDL 2026 🤖 ❄️
His tutorial will give you an understanding of the rationale behind non-Euclidean methods and prepare you to explore them for your own research problems.
More info in the comments 👇
His tutorial will give you an understanding of the rationale behind non-Euclidean methods and prepare you to explore them for your own research problems.
More info in the comments 👇
September 3, 2025 at 6:45 AM
We are thrilled to welcome Aiden Durrant as a Winter School speaker at NLDL 2026 🤖 ❄️
His tutorial will give you an understanding of the rationale behind non-Euclidean methods and prepare you to explore them for your own research problems.
More info in the comments 👇
His tutorial will give you an understanding of the rationale behind non-Euclidean methods and prepare you to explore them for your own research problems.
More info in the comments 👇
Reposted by Pascal Mettes
Many thanks @pascalmettes.bsky.social @ellisamsterdam.bsky.social for visiting us @bristoluni.bsky.social to examine (now Dr) Adriano Fragomeni (supervised by myself and Michael Wray) and give a great talk on hyperbolic deep learning. Enjoyed your visit
July 23, 2025 at 9:17 PM
Many thanks @pascalmettes.bsky.social @ellisamsterdam.bsky.social for visiting us @bristoluni.bsky.social to examine (now Dr) Adriano Fragomeni (supervised by myself and Michael Wray) and give a great talk on hyperbolic deep learning. Enjoyed your visit
We are back at #ICCV2025 for the second workshop on Hyperbolic and Hyperspherical Learning for Computer Vision.
We will have 2 tracks: one for new research (to be published in proceedings) and one for recently published works.
Deadline: June 1st
Link: sites.google.com/view/beyonde...
We will have 2 tracks: one for new research (to be published in proceedings) and one for recently published works.
Deadline: June 1st
Link: sites.google.com/view/beyonde...
Beyond Euclidean - Call for Papers
Important Dates:
Submission Portal Opens: 29th of April 2025
Submission Deadline: 1st of June 2025 (AOE)
Preliminary Author Notification Deadline: 27th of June 2025
Camera-ready deadline (Proceeding...
sites.google.com
May 1, 2025 at 6:21 AM
We are back at #ICCV2025 for the second workshop on Hyperbolic and Hyperspherical Learning for Computer Vision.
We will have 2 tracks: one for new research (to be published in proceedings) and one for recently published works.
Deadline: June 1st
Link: sites.google.com/view/beyonde...
We will have 2 tracks: one for new research (to be published in proceedings) and one for recently published works.
Deadline: June 1st
Link: sites.google.com/view/beyonde...
Today I will give a keynote on "Hyperbolic Visual Understanding" at the Non-Euclidean Foundation Models and Geometric Learning (NEGEL) workshop at the Web Conference in Sydney.
If you are at #WWW2025 and want to discuss hyperbolic learning, I'm around all week!
If you are at #WWW2025 and want to discuss hyperbolic learning, I'm around all week!
April 28, 2025 at 11:16 PM
Today I will give a keynote on "Hyperbolic Visual Understanding" at the Non-Euclidean Foundation Models and Geometric Learning (NEGEL) workshop at the Web Conference in Sydney.
If you are at #WWW2025 and want to discuss hyperbolic learning, I'm around all week!
If you are at #WWW2025 and want to discuss hyperbolic learning, I'm around all week!
Great to see you all at #ICLR2025, it was fun discussing our works on hyperbolic vision-language models, a better way to do object detection, and brain-aligned image generation.
See you at the next one!
See you at the next one!
April 28, 2025 at 11:14 PM
Great to see you all at #ICLR2025, it was fun discussing our works on hyperbolic vision-language models, a better way to do object detection, and brain-aligned image generation.
See you at the next one!
See you at the next one!
I am very excited to present at the Pattern Recognition and Computer Vision Colloquium at CTU Prague. I will be there Wednesday afternoon and Friday morning as well if you are around and want to discuss hyperbolic learning!
The Visual Recognition Group at CTU in Prague organizes the 49th Pattern Recognition and Computer Vision Colloquium with D. Karatzas, M. Masana, T. Tommasi, P. Mettes @pascalmettes.bsky.social , E. Brachmann @ericbrachmann.bsky.social and V. Stojnic @stojnicv.xyz
cmp.felk.cvut.cz/colloquium/#...
cmp.felk.cvut.cz/colloquium/#...
April 7, 2025 at 2:53 PM
I am very excited to present at the Pattern Recognition and Computer Vision Colloquium at CTU Prague. I will be there Wednesday afternoon and Friday morning as well if you are around and want to discuss hyperbolic learning!
Reposted by Pascal Mettes
Who's up for writing a CVF motion proposing to vary the location of CVPR?
It being forever in the US, with the concerning (euphemism) situation there, is a recipe for losing this amazing community.
But even without the current events, it should have been done ages ago.
It being forever in the US, with the concerning (euphemism) situation there, is a recipe for losing this amazing community.
But even without the current events, it should have been done ages ago.
March 11, 2025 at 7:59 AM
Who's up for writing a CVF motion proposing to vary the location of CVPR?
It being forever in the US, with the concerning (euphemism) situation there, is a recipe for losing this amazing community.
But even without the current events, it should have been done ages ago.
It being forever in the US, with the concerning (euphemism) situation there, is a recipe for losing this amazing community.
But even without the current events, it should have been done ages ago.
Vision-language models are hierarchical and asymmetrical, hence all models benefit from hyperbolic embeddings.
Our new #ICLR2025 oral paper shows how image-box compositions create cool new hierarchies and powerful vision-language models when used during training!
Link: arxiv.org/abs/2410.06912
Our new #ICLR2025 oral paper shows how image-box compositions create cool new hierarchies and powerful vision-language models when used during training!
Link: arxiv.org/abs/2410.06912
March 4, 2025 at 9:48 AM
Vision-language models are hierarchical and asymmetrical, hence all models benefit from hyperbolic embeddings.
Our new #ICLR2025 oral paper shows how image-box compositions create cool new hierarchies and powerful vision-language models when used during training!
Link: arxiv.org/abs/2410.06912
Our new #ICLR2025 oral paper shows how image-box compositions create cool new hierarchies and powerful vision-language models when used during training!
Link: arxiv.org/abs/2410.06912
Reposted by Pascal Mettes
"Compositional Entailment Learning for Hyperbolic Vision-Language Models" #ICLR25
With: Avik Pal, Max van Spengler, Guido Maria D'Amely di Melendugno, Alessandro Flaborea, and @pascalmettes.bsky.social
Paper: arxiv.org/abs/2410.06912
With: Avik Pal, Max van Spengler, Guido Maria D'Amely di Melendugno, Alessandro Flaborea, and @pascalmettes.bsky.social
Paper: arxiv.org/abs/2410.06912
January 25, 2025 at 1:17 PM
"Compositional Entailment Learning for Hyperbolic Vision-Language Models" #ICLR25
With: Avik Pal, Max van Spengler, Guido Maria D'Amely di Melendugno, Alessandro Flaborea, and @pascalmettes.bsky.social
Paper: arxiv.org/abs/2410.06912
With: Avik Pal, Max van Spengler, Guido Maria D'Amely di Melendugno, Alessandro Flaborea, and @pascalmettes.bsky.social
Paper: arxiv.org/abs/2410.06912
Hyperbolic learning is growing rapidly by the day. From weekly alerts in 2023 to daily digests in 2024!
From our current research, it is clear that 2025 will be a huge year for hyperbolic learning research.
I had an interview to elaborate our research:
shorturl.at/CQD53
From our current research, it is clear that 2025 will be a huge year for hyperbolic learning research.
I had an interview to elaborate our research:
shorturl.at/CQD53
A deep dive into mathematics for more accurate AI
Deep learning networks underpin many of today's AI applications, such as image recognition. Those networks can still fail massively when they need to name an object, such as an animal. According to Uv...
shorturl.at
December 16, 2024 at 9:46 AM
Hyperbolic learning is growing rapidly by the day. From weekly alerts in 2023 to daily digests in 2024!
From our current research, it is clear that 2025 will be a huge year for hyperbolic learning research.
I had an interview to elaborate our research:
shorturl.at/CQD53
From our current research, it is clear that 2025 will be a huge year for hyperbolic learning research.
I had an interview to elaborate our research:
shorturl.at/CQD53